FuzzBench is an open source platform for rigorously evaluating fuzzers on a wide variety of real-world benchmarks. The goal of FuzzBench is to make it painless to evaluate fuzzing research and make fuzzing research easier for the community to adopt.

FuzzBench provides:

• An easy API for integrating fuzzers.
• Benchmarks from real-world projects. FuzzBench can use any OSS-Fuzz project as a benchmark.
• A reporting library that produces reports with graphs and statistical tests to help you understand the significance of results.

This repository is a slim, local-only fork focused on running experiments with Docker on your own machine.

See docs/running-a-local-experiment/ for setup and configuration.

After integrating a fuzzer, follow docs/getting-started/ to build and test it, then run an experiment with experiment/run_experiment.py.

Reports are written to the report_filestore path in your experiment config (for example /tmp/report-data/$EXPERIMENT_NAME/index.html).

You can view a sample report and periodically generated reports from past upstream FuzzBench experiments. The sample report uses 10 fuzzers against 24 real-world benchmarks, with 20 trials each over 24 hours.

When analyzing reports, we recommend:

• Checking the strengths and weaknesses of a fuzzer against various benchmarks.
• Looking at aggregate results to understand the overall significance of the result.

Please provide feedback by opening a GitHub issue here.

See the docs/ directory (or serve it with make docs-serve) for the full documentation set.

This fork is maintained at github.com/Mrmaxmeier/fuzzbench. The upstream project mailing list is fuzzbench-users.